Provide spans of named entities

[caufieldjh]

Where possible (i.e., an extracted named entity is explicitly mentioned in the input text), its corresponding span will be provided in the list of named entities, a la:

named_entities:
  - id: AUTO:erosion
    label: erosion
    original_spans:
      - 29:35
  - id: SNOMEDCT:263913002
    label: ulceration
    original_spans:
      - 44:53

These will not be included if the named entity isn't present in the input text, and this may happen if the prompt instructs the raw response to be transformed in some way (e.g., the raw text contains "ulceration" but the extraction instructions cause the LLM to represent it as "TRUE", as in "yes there is an ulceration", then the span details won't appear)